1. The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology Chris Lucas, Hanh Nguyen Bureau of Meteorology – R&D Branch Regional characteristics of tropical expansion and the role of climate variability

2. Observations vs Models Observed rate in many metrics is 0.5 – 1.0 degrees/decade of expansion. (e.g. Davis and Rosenlof 2012; Lucas et al 2014) Rate expected in climate change projections (CMIP3, CMIP5) is typically 0.1-0.2 degrees per decade Why? Several hypotheses… Model deficiencies – sensitivity too low, or simply incorrect Observational issues Short data sets (most from 1979) Reanalysis problems – inhomogeneities, poor constraint by observations Choice of metric – different metrics, different responses in some cases Use Tropopause Height Frequency methods based on actual radiosonde data. No Reanalyses! 2

3. Data and methodology Focus here is on using real observations (not reanalyses) to examine tropical expansion Basis for evaluation of reanalyses Data used: Integrated Global Radiosonde Archive (IGRA) from NOAA Comprehensive global coverage, freely available from web Methodology thoroughly described in Lucas et al (2012, JGR) Use all available data Create time/latitude array of tropical tropopause days (TTD) in ‘bands’ ‘First difference’ compositing technique Sampling bias corrections, error estimation Contour time/latitude array to estimate trends, use TTD=300,200,100 and 50

4. Summary of Lucas et al (2012) Analysis of SH expansion 3 regions (ANZ, AFR, SA) + average Focus on TTD=200 contour Comparison with four reanalyses NCEP, NCEP2, ERA-40, ERA-I Same methodology applied Reanalysis contours shifted poleward Trends (SH only) sondes: 0.4 deg dec -1 (expansion) NCEP, NCEP2: 0.3 – 0.5 deg dec -1 ERA-I: no trend Two periods of notable difference post-2002 -- better satellite observations improving ERA-I, creates inhomogeneity pre-1985 – ?? Repeat this methodology for the NH, compare with SH results

5. Study regions illustrated Region Bands Sites ANZ 10 45 AFR 8 26 SA 9 36 ASIA 21 252 EUR 19 178 NA 20 136

6. NH TTD contours by region Structure of ‘subtropics’ different in the regions of the NH Less poleward extent in EUR 300,200 contours shifted poleward in ASIA Thickest in NA Significantly different variability in NA ‘Dips’ on 300 contour Responses around 2000 Volcanic response? Trends (since 1979) Largest in ASIA (0.5 - 0.8) Insignificant in NA (0-0.3) Moderate in EUR (0.4-0.5) Numbers refer to band locations

7. TTD contours from ASIA 300 200 100 50 black=sonde coloured=reanalysis Good match for interannual variability General shift poleward of reanalysis contours 1.5-3.0 degrees here Less in NA, EUR Improvement in ERA-I shift (red lines) at end of period

8. NH ‘global’ summary Weighted average TTD=200 contour across all regions Removing mean position accounts for shift Volcanic response more visible in this view Generally good agreement prior to 2002 1987-88? Significant differences occur after 2002, just as for SH Suggests inhomogeneity in reanalysis fields Hypothesis: related to significant improvement in satellite instrumentation (AIRS) BUT… There appears to be little sign of this poleward of 35 N…data match up very well there

9. NH and SH Tropical Expansion 9 ContourNHSH 3000.42 (0.20)0.32 (0.29) 2000.29 (0.15)0.45 (0.23) 1000.38 (0.20)0.57 (0.26) 500.52 (0.22)0.49 (0.38) SH expansion greater than NH expansion, but difference not statistically significant

10. Partial Correlation and Regression Examine role of natural climate variability in tropical expansion using partial correlation and regression techniques Isolate the effects of a single variable Seven modes: ENSO, PDO, PNA, SAM, AO, NAO, AMO Detrend all variables Only interannual effects captured Relate annual global and regional TTD contours to annual and seasonal time series of mode time series The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology 10

11. Results ENSO matters almost everywhere Expansion during La Nina ‘Reverse’ effect in NA Regression mag: 0.2-0.5 deg/unit, more on ‘extratropical’ JJA, SON PDO mostly in ASIA and NH Shifts whole tropics toward north during negative phase Regression mag: 0.3-0.5 deg/unit All seasons but MAM SAM strongly affects ANZ during MAM Weak general effects in SH PNA, AMO show regional, specific interannual effects AO, NAO results confusing, no strong effects noted The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology 11

12. 1-1.5 units of change Estimating effect on TE trends Previous effects were interannual only Many mode indices have apparent trends over 1979-2012 period ENSO, PDO, SAM, AMO Estimate number of units that index changes Multiply by regression coefficient to get amount of change, Divide by time to get average trend PDO Index 1.5 units * 0.6 deg/unit = 0.9 deg 0.9 deg / 3.3 decades = 0.27 deg dec -1 Up to 50% of observed trend

13. Regional Tropical Expansion and Climate Variability 13 0.1 - 0.5 0.5 - 0.80.2 - 0.3 0.3 - 0.8 0.4 – 0.5 Global: ENSO 10-30% (except *) PDO up to 50% SAM 20-30% * *

14. Discussion Regional variability/effects key to understanding impacts of TE, particularly in NH Possible Mechanisms Interannual (e.g ENSO, SAM) – shifts in storm tracks, eddy jet  changes in baroclinic waves and shifts in tropical edge…eg. Chen and Held (2007), Chen et al (2008), Ceppi and Hartmann (2013) Decadal (e.g. PDO, AMO) – changes to SST patterns eg. Staten et al (2012), Quan et al (2014), anomalous heating/cooling of midlatitudes from SST pattern (eg. Allen et al 2012) Apparent tropical expansion since 1979 has been enhanced by natural variability Climate models won’t necessarily capture this variability. What happens when modes shift (e.g. PDO, ‘end’ of ozone hole)? The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology 14

15. The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology Your name Your details Thank you.

16. NH/SH ‘global’ comparison Subtropics in NH are larger compared to SH Start in same place, but extend further poleward Likely related to greater land are in NH Analogous to finding with other variables (e.g.  ) Is tropical expansion asymmetric? Trends in SH on 300,100, 50 less reliable (data issues) SH trends are larger on 200,100 contours, but not statistically significantly so (about 1-  difference)

17. Observations of Tropical Expansion Tropical Expansion is real Observed across multiple datasets and a variety of metrics Consensus of observations suggests an expansion rate of 0.5-1.0 degrees/decade from 1979 Expansion is observed in both hemispheres From Lucas et al. (2014)

18. 18